You are overpaying for… | The AI Corner Summary

1. Key Themes

Theme 1: The AI Model Wars Have Become a Pricing War

Grok 4.5's launch signals a structural shift from capability competition to cost competition at the frontier level.

"Yesterday, the model wars turned into a pricing war. SpaceXAI shipped Grok 4.5, its first model trained alongside Cursor, at $2 per million input tokens and $6 per million output."

The cost delta across equivalent tasks is now enormous: "running the same coding-agent benchmark task costs $2.49 in Grok Build, $5.07 in Codex, and $11.80 with Fable 5 in Claude Code. Same job. Five times the bill."

Theme 2: Token Pricing Is a Misleading Metric — Task-Level Economics Are What Matter

The industry has trained buyers to optimize for the wrong unit of measurement.

"Stop pricing models per token. Price them per completed task. Per-token prices lie. Sonnet 5 costs a fifth of Opus 4.8 per token, and at high effort it costs more per finished task, because it talks more."

The key driver: "Grok 4.5 wins on economics because it uses 4.2x fewer output tokens than Opus on the same engineering tasks. Verbosity is a price. Almost everyone ignores it."

Theme 3: AI Spend Is Ballooning Into a Material Budget Problem for Engineering Teams

This is no longer a marginal line item — it's becoming a significant operational cost center.

"The average team pays $150 to $250 per developer per month for AI, and Uber's engineers burned their entire 2026 AI budget in four months at $500 to $2,000 a head."

The opportunity: "Routing typically cuts that spend 40 to 70%."

Theme 4: Model Routing Is Emerging as a Distinct, High-Value Discipline

Choosing the right model for the right task — not just picking the best model — is the new optimization layer.

"The question is no longer 'which model is best.' It is 'which model is best for this task at this price,' and that question has a table for an answer."

The routing framework spans task type, effort level, escalation logic, and re-route triggers across "Grok 4.5, Fable 5, Opus 4.8, Sonnet 5, GPT-5.5, Gemini, Haiku, and the open models."

2. Contrarian Perspectives

Contrarian 1: A "Cheaper" Model Per Token Can Be More Expensive Per Task Than a "Premium" Model

The consensus is that lower token prices = lower costs. The article directly dismantles this.

"Sonnet 5 costs a fifth of Opus 4.8 per token, and at high effort it costs more per finished task, because it talks more."

This is a trap that most per-token shoppers fall into — verbose models rack up output costs that erase their sticker-price advantage.

Contrarian 2: Effort Level Matters More Than Model Choice

Buyers fixate on model selection, but the "effort dial" is actually the bigger cost lever.

"The effort dial, the single setting that swings cost 6x, triple the gap between any two model prices."

This implies that misconfiguring effort on a cheap model can cost more than using a premium model at the right effort setting — a counterintuitive operational insight most teams miss.

Contrarian 3: Grok 4.5 Is a Frontier-Tier Model, Not a Budget Alternative

The market assumption is that low-cost models sacrifice quality. The benchmark data challenges this.

"Grok 4.5 ranks #4 on the Artificial Analysis Intelligence Index, behind only Fable 5, GPT-5.5, and Opus 4.8. And it hits those scores at a fraction of the spend: $0.31 per task on the index, and $0.49 per task on GDPval knowledge work, roughly 90% cheaper than the models ranked above it."

Musk's own framing reinforces this: "It is an Opus-class model, but faster, more token-efficient and lower cost."

3. Companies Identified

xAI (SpaceX AI)

Description: Elon Musk's AI division
Why mentioned: Released Grok 4.5, the central subject of the article — a frontier-tier model at dramatically lower per-task cost
Quote: "SpaceXAI shipped Grok 4.5, its first model trained alongside Cursor, at $2 per million input tokens and $6 per million output."

Cursor

Description: AI-powered coding environment
Why mentioned: Co-trained with Grok 4.5; also mentioned as doubling usage limits during the free window
Quote: "Grok 4.5, its first model trained alongside Cursor... Cursor doubles your usage."

Uber

Description: Ride-sharing / technology company
Why mentioned: Used as a real-world case study of runaway AI spend
Quote: "Uber's engineers burned their entire 2026 AI budget in four months at $500 to $2,000 a head."

Anthropic (implied via Opus 4.8, Sonnet 5, Claude Code)

Description: AI safety company and frontier model provider
Why mentioned: Benchmarked as the high-cost incumbent; Sonnet 5 cited as the "trap" model; Claude Code cited as most expensive coding agent
Quote: "$11.80 with Fable 5 in Claude Code. Same job. Five times the bill."

OpenAI (implied via Codex, GPT-5.5)

Description: Leading AI lab
Why mentioned: Included in routing comparison; Codex cited as mid-tier cost option for coding tasks
Quote: "$5.07 in Codex" vs. "$2.49 in Grok Build."

4. People Identified

Elon Musk

Description: CEO of xAI / SpaceX / Tesla
Why mentioned: Quoted directly on Grok 4.5's positioning
Quote: "It is an Opus-class model, but faster, more token-efficient and lower cost."

Ruben Dominguez

Description: Author of The AI Corner newsletter
Why mentioned: Author of the article and creator of the AI Model Router framework
Quote: "I built it" — referring to the routing table mapping every task type to the optimal cost-per-task model.

5. Operating Insights

Insight 1: Measure AI costs per completed task, not per token — immediately

Switching the unit of measurement reveals hidden overcharges that per-token pricing obscures. The benchmark data is already available to make this switch today.

"Stop pricing models per token. Price them per completed task. Per-token prices lie."

Insight 2: Audit your effort dial before changing models

Before switching model providers, operators should check their effort setting — it has more cost leverage than model selection itself.

"The effort dial, the single setting that swings cost 6x, triple the gap between any two model prices."

Insight 3: Use the free window now

There is a time-limited arbitrage opportunity: Grok 4.5 is currently free in Grok Build, and Cursor is doubling usage limits. Run your benchmark tasks now to validate fit before committing budget.

"The free-window plays, what to run this week while Grok 4.5 costs zero in Grok Build and Cursor doubles your usage."

6. Overlooked Insights

Overlooked Insight 1: The "Cursor Code-Custody Problem"

The article flags a "Cursor code-custody problem" as one of three warnings about Grok 4.5 — but gives no detail in the free portion. For teams using Cursor as their primary dev environment, this is potentially a significant IP or data governance risk worth investigating before adoption.

"The three warnings, the quality dispute, the Cursor code-custody problem, and the regulatory overhang, priced honestly."

Overlooked Insight 2: The Escalation Ladder — 90%+ of Frontier Power at Half the Cost

The article hints at a specific pattern for accessing top-tier model quality without paying full freight — but reserves the details for subscribers.

"The escalation ladder, when to route up to Fable 5, and the two patterns that keep 90%+ of its power at half the cost."

This suggests there are structured prompting or workflow patterns that extract near-frontier performance from cheaper models — a meaningful finding for teams currently defaulting to the top-tier model for all tasks.